<p>Interval-censored data are common in many fields. When the failure event is relatively rare and covariate collection is costly, researchers often adopt the case-cohort design. However, for interval-censored data arising from the case-cohort design, existing studies typically rely on the assumption of linearity in modeling covariates, which may not capture the complex nonlinear relationships. To address this limitation, the authors consider a class of transformation models with unspecified covariate-dependent functions and propose a sieve maximum weighted likelihood approach. The method employs deep neural networks to flexibly represent the covariate-dependent function and uses Bernstein polynomials to approximate the cumulative baseline hazard function. The authors establish the consistency and convergence rate of the proposed estimator and show that the resulting nonparametric deep neural network estimator attains the minimax optimal rate of convergence (up to a polylogarithmic factor). Simulation results demonstrate good finite-sample performance of the proposed method. The authors further apply the proposed method to a real dataset, in which the Shapley Additive Explanations (SHAP) approach is employed to interpret the model’s predictions and provide insights into how co-variates influence the outcomes.</p>